And What It Would Take to Replicate It Elsewhere
Foreword — by 
This post is a direct follow up to my data analysis post of the #introduceyourself tag. There, the standout finding was the impact of Spanish language and possibly language density on user retention. This deserves its own deeper dive, so once again I have worked with Claude to dig into what the data tells us.
The result: there is a lot of potential for improved growth that we are not yet exploring like we should.
Update (2026-05-07): This post has been corrected based on feedback from @danielvehe. 
Research Findings — by Claude Opus 4.6
This analysis was conducted at
The Starting Point: A 14pp Retention Gap
The #introduceyourself analysis found that Spanish-tagged first-posts retain at roughly double the rate of non-Spanish first-posts — consistently, across every cohort year from 2017 to 2023. That post identified the gap but didn't decompose it. Three plausible explanations presented themselves:
- Language community — Spanish-speaking newcomers land in a dense, in-language subcommunity that welcomes and supports them.
- Economic motivation — many Spanish-speaking Hivers are Venezuelan, and a few dollars in HBD can be life-altering in a hyperinflationary economy.
- Venezuelan-specific culture — maybe this isn't a "Spanish" effect at all, but something unique to Venezuela's relationship with this blockchain.
These aren't mutually exclusive. The challenge is separating them.
Investigation 1: Is It Venezuelan, Not Hispanic?
Country-level retention
Using profile.location fields from the Accounts table (more reliable than post-level country tags, which only 27% of Spanish first-posters use), I split the 2023 Spanish-tagged cohort by country:
| Country | Spanish-tagged | n | Retention |
|---|---|---|---|
| Cuba | Yes | 95 | 55.8% |
| Venezuela | Yes | 414 | 45.4% |
| Argentina | Yes | 26 | 42.3% |
| Other location | Yes | 170 | 42.9% |
| No location set | Yes | 553 | 21.9% |
For comparison, the non-Spanish baseline:
| Country | Spanish-tagged | n | Retention |
|---|---|---|---|
| Venezuela | No | 348 | 19.8% |
| No location | No | 4,758 | 13.9% |
| Other location | No | 3,347 | 27.0% |
Two findings jump out immediately.
Finding 1: Cuba leads, not Venezuela. Cuban Spanish-tagged newcomers retain at 55.8% — 10 percentage points above Venezuela. Cuba only appeared at scale in the 2022 cohort (n=75, 40%) and grew further in 2023. This is a new and fast-growing community, not just a Venezuelan spillover.
Finding 2: Non-Spanish Venezuelans retain at baseline. Venezuelans who don't use the Spanish tag retain at 19.8% — statistically indistinguishable from the 19.1% unclassified baseline. Same country, same economic pressure, but without community membership, no retention advantage.
This second finding is the critical one. It rules out pure economic motivation as the explanation. If Venezuelans stayed because HBD matters in a hyperinflationary economy, all Venezuelans would retain at elevated rates, regardless of whether they tagged their posts in Spanish. They don't. The retention boost requires community membership.
But community membership alone doesn't explain the full picture either
Consider: Spanish-tagged users with no profile location — who may or may not be Venezuelan — retain at only 21.9%. That's barely above the non-Spanish baseline. If in-language community were the whole story, any Spanish-tagged user should benefit, regardless of where they're from. The high retention seems to require both community membership and something else.
That "something else" starts to look like economic motivation when we check the longitudinal data.
The time dimension
To test whether the retention advantage is a recent phenomenon or something structural, I pulled yearly cohorts from 2017 through 2022:
| Year | VE + Spanish | Retention | Non-Spanish baseline |
|---|---|---|---|
| 2017 | n=1,945 | 54.6% | 18.4% |
| 2018 | n=6,053 | 32.3% | 16.0% |
| 2019 | n=770 | 14.5% | 14.6% |
| 2020 | n=555 | 46.9% | 10.5% |
| 2021 | n=1,266 | 47.9% | 14.5% |
| 2022 | n=559 | 43.7% | 14.3% |
The 2019 collapse is telling. During the crypto winter, Venezuelan+Spanish retention crashed to 14.5% — matching the non-Spanish baseline exactly. When HBD lost most of its purchasing power, the economic lever stopped working, and community membership alone wasn't enough to keep people.
Note also: 2017 retention was 54.6%, before Aliento was founded (March 2018). The community density effect predates Aliento's curation infrastructure. As 
Three Stacked Effects
The pattern across these cuts points to three separable retention effects:
1. Profile commitment (+6pp). Users who set any profile location — regardless of country — retain ~6 percentage points higher than those with blank profiles. This is an effort signal: the act of filling out your profile correlates with higher baseline engagement. It's not causal in the way the other effects are, but it's a confounder we need to account for.
2. Economic motivation (+11pp). Among users with profiles set, Venezuelans without Spanish tags retain at 19.8%. Venezuelans with Spanish tags retain at 45.4%. But the non-Spanish Venezuelan rate (19.8%) is itself elevated above the no-location baseline (13.9%) by about 6pp — that's the profile commitment effect. The economic motivation effect is visible in how non-Spanish Venezuelans compare to, say, non-Spanish Americans (25.4%) or non-Spanish Brits (24.6%). It's there, but it's modest without community membership.
3. In-language curation infrastructure (+10pp minimum). The gap between non-Spanish Venezuelans (19.8%) and Spanish-tagged Venezuelans (45.4%) — a 26pp difference that can't be explained by economics alone, since both groups face identical economic conditions — represents the combined effect of community membership and the curation infrastructure that comes with it. After accounting for profile commitment, the curation infrastructure adds roughly 10-20 percentage points.
The 2019 crypto winter provides the natural experiment that separates these effects. When the economic lever collapsed, the curation infrastructure effect collapsed with it — suggesting the two are multiplicative, not additive. Curation infrastructure amplifies economic motivation; without something worth staying for economically, even a welcoming community isn't enough.
Investigation 2: A Global Natural Experiment
If the retention advantage requires both economic motivation and community infrastructure, we should be able to find natural control groups: countries with economic motivation but no community infrastructure, and countries with neither.
Multi-country retention
I pulled retention rates for all identifiable countries, pooling the 2021-2023 cohorts for sample size:
| Country | n | Retention | Economic pressure? | Community infrastructure? |
|---|---|---|---|---|
| Cuba + Spanish | 187 | 49.2% | High | Yes (Aliento + HiveCuba) |
| Venezuela + Spanish | 2,263 | 46.4% | High | Yes (Aliento, Rutablockchain) |
| Ghana | 76 | 47.4% | High | Unclear (small n) |
| Ukraine | 111 | 46.9% | High (war) | Unclear |
| Mexico + Spanish | 39 | 46.2% | Moderate | Yes (Aliento) |
| Spain + Spanish | 58 | 55.2% | Low | Yes (Aliento) |
| Argentina + Spanish | 94 | 41.5% | High | Yes (Aliento) |
| Indonesia | 843 | 37.4% | High | Partial (legacy Steemit community) |
| Nigeria | 1,527 | 36.0% | High | No |
| Philippines | 862 | 35.3% | High | Partial (Cebu teachers) |
| Bangladesh | 342 | 35.4% | High | No |
| India | 505 | 31.5% | Moderate | No |
| USA | 820 | 25.4% | Low | No |
| UK | 203 | 24.6% | Low | No |
| Unclassified, non-Spanish | 35,394 | 19.1% | Mixed | No |
The 2×2 resolves cleanly:
- Economic pressure + community infrastructure (Venezuela, Cuba, Mexico, Argentina + Spanish): 41–55% retention
- Economic pressure + no community infrastructure (Nigeria, Philippines, Bangladesh): 35–37% retention
- No economic pressure + no infrastructure (USA, UK): 24–25% retention
- No economic pressure + community infrastructure (Spain + Spanish): 55% — the most interesting cell
Spain + Spanish at 55.2% retention (n=58, small but consistent across years) suggests that community infrastructure can deliver high retention even without economic pressure. That's a small sample, but it's the right direction for the hypothesis.
Nigeria's 36% — achieved with zero community infrastructure — shows how powerful economic motivation alone can be. Nigerian newcomers retain at nearly double the rate of Americans, despite receiving far less curation support (as we'll see below). If Nigeria got even basic community infrastructure, the data suggests retention could climb significantly.
Investigation 3: What Does "Community Infrastructure" Actually Look Like?
To understand what separates these tiers, I examined the curation infrastructure serving each country's newcomers. I pulled the top voters and commenters on first-posts from the 2023 cohort for Venezuela+Spanish (via Aliento), Nigeria, and the Philippines.
Aliento: A full-stack curation community
The top voters on Aliento community (hive-110011) first-posts in 2023:
The curator team is pan-Hispanic, not purely Venezuelan: Venezuelan, Mexican, Colombian, and Cuban curators all appear in the top 50. 
This matters: it means Aliento's curation infrastructure serves the entire Spanish-speaking community, not just Venezuelans. A Cuban newcomer lands in the same curation ecosystem as a Venezuelan or Argentine one.
(Correction: an earlier version of this post listed
Nigeria: Zero community infrastructure
The top voters on Nigerian first-posts:
| Voter | Votes | Location |
|---|---|---|
 | 153 | Netherlands ( ) |
 | 128 | — |
 | 119 | — |
| 116 | Welcome initiative (Netherlands) |
 | 110 | — |
 | 110 | — |
Not a single Nigerian-located account appears in the top 50 voters on Nigerian first-posts. The votes come from
Philippines: The comment layer
The Philippines presented a puzzle. Filipino first-posts received more average votes (284) than Venezuelan+Spanish ones (219). But when I pulled the voter list, the picture was different:
The top voters were 
Below the bots and generic welcomers, a Cebu-based network of teachers actively comments on every Filipino newcomer's first post. 
There's also 
Three tiers of infrastructure
This analysis reveals three distinct tiers:
The Philippines has organic community commenting but outsources its economic signal to OCD's generic curation trail. Aliento provides both: dedicated voting (economic signal that says "your community values this") and personal commenting (social signal that says "we see you and welcome you").
The interesting puzzle: Nigeria retains at 36%, matching the Philippines' 35%, despite having no community infrastructure at all. This is the economic motivation effect in isolation — Nigerian newcomers stay because HBD has real purchasing power, even without anyone welcoming them. If Nigeria developed even the Filipino-level comment infrastructure, the data suggests retention could climb meaningfully. If it developed Aliento-level full-stack curation, the jump could be dramatic.
Investigation 4: Engagement Density by Country
To quantify the welcome experience, I measured engagement density on first-posts across countries: average distinct voters, average distinct commenters, time to first comment, and percentage of first-posts that received any comment at all.
| Country | Retained? | n | Avg voters | Avg commenters | First comment (min) | % got comments |
|---|---|---|---|---|---|---|
| VE + Spanish | Yes | 185 | 219 | 10.4 | 112 | 98% |
| VE + Spanish | No | 226 | 147 | 7.1 | 299 | 90% |
| Philippines | Yes | 94 | 284 | 13.5 | 180 | 87% |
| Philippines | No | 179 | 144 | 8.0 | 1,798 | 87% |
| Nigeria | Yes | 114 | 97 | 8.0 | 248 | 92% |
| Nigeria | No | 267 | 61 | 4.9 | 526 | 79% |
| Indonesia | Yes | 24 | 60 | 4.8 | 350 | 88% |
| Indonesia | No | 69 | 12 | 1.9 | 631 | 58% |
| USA | Yes | 34 | 56 | 5.2 | 1,161 | 74% |
| USA | No | 140 | 19 | 1.8 | 2,809 | 42% |
The time-to-first-comment metric tells the story most clearly. A Venezuelan+Spanish newcomer who will later be retained gets their first comment in 112 minutes — under two hours. A Filipino newcomer waits 180 minutes. A Nigerian newcomer waits 248 minutes. An American newcomer who will not be retained waits 2,809 minutes — nearly two days.
For non-retained Americans, 42% received zero comments at all. For non-retained Venezuelans in the Spanish community, only 10% received zero comments. The welcome experience is categorically different.
The Filipino vote counts (284 average) are higher than Venezuela's (219), but this is trail inflation — 120+ accounts auto-following OCD's vote. The quality of that engagement is different from 10 personal curators each evaluating your post. A trail vote is an economic signal without a social one. Aliento's votes carry both.
Investigation 5: The Unplugged Opportunity
In the 2023 cohort, I found roughly 280 Venezuelan first-posters who didn't tag their posts with spanish, espanol, or related tags. Of these:
- 107 posted on Liketu (hive-147010), of which 57 had likely Spanish-language body text
- 39 posted on DBuzz (hive-193084)
- 18 were already in the Aliento community despite not using the Spanish tag
- Roughly 150 per year are addressable: they're posting Spanish content through non-Spanish-tagged channels
These users are currently invisible to the Spanish curation infrastructure. A simple automated detection — checking for Spanish-language body content combined with Venezuelan profile location — could route them toward Aliento's curation pipeline at near-zero marginal cost. At the observed retention difference between Spanish-tagged and non-Spanish-tagged Venezuelans (45.4% vs. 19.8%), routing even half of these users could retain an additional ~19 users per year from a population that currently retains at baseline.
That's a small absolute number, but the cost is essentially zero — it's a detection and routing problem, not a resource allocation one.
What This Means for Hive
The three levers
This analysis identifies three retention levers with very different cost/benefit profiles:
Lever 1: Route unplugged Hispanics. ~150 addressable users/year posting Spanish content without Spanish tags. Near-zero cost — requires automated language/location detection and routing to existing Aliento infrastructure. Expected yield: ~19 additional retained users/year. This is the cheapest intervention.
Lever 2: Saturate the existing Hispanic pipeline. Aliento's curation infrastructure serves roughly 800-1,000 Spanish first-posters per year (those who tag and are in-community). But there's headroom — 553 Spanish-tagged users with no profile location retain at only 21.9%, suggesting they're tagged but not fully plugged in. Deepening community engagement for this group (Discord outreach, faster comment response, stronger curation coverage) could lift their retention closer to the 45% observed for location-identified users. Moderate cost, moderate yield.
Lever 3: Build infrastructure for a new community. Nigeria is the most obvious candidate: 1,527 first-posters pooled across 2021-2023, 36% retention with zero community infrastructure, and an English-speaking population that eliminates the language barrier. But building a "Nigerian Aliento" from scratch — finding and training community curators, establishing a Discord/Telegram presence, securing delegation for curation votes — is a high-cost, uncertain-outcome project. The Filipino experience shows that organic community formation is possible (the Cebu teacher network emerged without any formal project), but it took years and still lacks the voting infrastructure that makes Aliento effective.
The broader lesson
The deepest finding in this analysis is structural: retention on Hive is not primarily an individual-level phenomenon. It's a community-level one. The difference between 19% and 46% retention isn't explained by individual user quality, content quality, or even reward levels (though all of those matter). It's explained by whether the newcomer lands in a community that has built the infrastructure to welcome, support, and retain them.
This has implications beyond the Spanish-speaking community. Every subcommunity on Hive — whether defined by language, geography, topic, or platform — faces the same structural question: do they have the curation density to retain newcomers? The data from this analysis suggests that the threshold for effective retention infrastructure is not impossibly high. It requires:
- A small team of dedicated personal curators (Aliento operates with roughly 15-20 active curators across multiple countries)
- Fast comment response (under 2 hours for the best-performing community)
- Consistent voting coverage (Aliento's distinguishing feature from the intro-tag analysis was systematic floor-level support: 81-85% of newcomers receiving at least $1)
- Economic motivation that makes small rewards meaningful (the 2019 crypto-winter collapse shows this matters, though Spain's high retention suggests it's not strictly necessary)
The cheapest intervention is Lever 1: route the users who are already posting Spanish content but aren't connected to the curation infrastructure. The most ambitious is Lever 3: replicate what Aliento has built for a new community. The evidence suggests both would work — the question is whether anyone is willing to build it.
Caveats
- Profile location is self-reported and incomplete. Only ~40% of accounts have a usable location string. Users with no location may differ systematically from those who fill it in, which is why the "no location" group retains at lower rates across all categories. This is an effort-signal confound, not a geographic one.
- Country classification uses LIKE-matching on location strings. Some matches are noisy (e.g., "usa" appears in "Maracaibusano"). We validated against a vocabulary of ~8,000 literal location strings categorized by region, but some misclassification is inevitable.
- Small samples for several countries. Cuba (n=95 in 2023), Spain (n=58 pooled), Argentina (n=94 pooled), and several others have sample sizes where individual users meaningfully shift the retention percentage. The main comparisons (Venezuela n=2,263, Nigeria n=1,527, Philippines n=862) are robust.
- Engagement density is correlation, not causation. Users who receive 10 comments may have written better introductions. The engagement metrics describe the environment a newcomer lands in, not necessarily the cause of their retention. However, the cross-country variation — where the same platform produces radically different engagement environments for different communities — argues against a pure selection story.
- "Retention" is defined as any top-level post 90-180 days after account creation. This captures medium-term engagement but not long-term commitment. A user who posts once at day 91 and never again counts as "retained." The relative comparisons between groups are still valid, but the absolute retention percentages overstate long-term engagement.
- The "extraction critique" — whether economically distressed user communities create or extract net value — is not addressed in this analysis. Prior work on Granger causality between user activity and price and the rewards-retention gradient suggests that user retention is net-positive for the network, but a direct cost/benefit analysis of community-specific value creation versus reward-pool draw would strengthen the argument. That's a topic for future investigation.
- OCD trail inflation means raw vote counts are misleading as a measure of community support. The Philippines' 284 average voters versus Venezuela's 219 would suggest stronger Filipino curation — but the Filipino votes come from an auto-trail, not personal evaluation. Commenter counts and time-to-first-comment are more honest measures of community engagement.
Correction (2026-05-09): An earlier version of this post described
Data: HiveSQL, queried May 2026. Cohorts: 2017-2023 account creation years, with retention measured as any top-level post 90-180 days after account creation. Country classification by profile.location LIKE-matching. Language classification by json_metadata tag detection (spanish, espanol, español, esp). Queries available on request. For the full retention series, see